CIAO Jonathan Mc Dowell Science Data Systems CUC

CIAO Jonathan Mc. Dowell Science Data Systems CUC Oct 2011 CXC-SDS

Science Data Systems activities - update Data Systems activities SDS develops requirements and algorithms for the CXCDS (Data Systems) team, then tests and documents the resulting software and provides user support. SDS is the interface between the analysis software and the user community; most of the work I discuss is done in collaboration with CXCDS SDS primary focus is now on maintenance and user support What we do Drivers (why we do it) 1 Maintenance & User Support (releases, docs. . ) 40% Internal (plus user input) 2 Making Chandra analysis easier 15% CUC, helpdesk, users 3 Instrument/observatory/Calibration changes 15% CXC Cal Team 4 Improving algorithms and data quality 10% Review of bugs, analyse user issues 5 Catalog: maximize science legacy 15% 6 New advanced techniques 5% Led by CXCDS (Ian's talk) Internal, and input from advanced users

1 Maintenance/User Support • • • 1. 1 1. 2 1. 3 1. 4 1. 5 1. 6 CUC Oct 2011 CIAO 4. 3 downloads Releases Community outreach and support Helpdesk Documentation Sherpa improvements CXC-SDS

CIAO 4. 3 Downloads CIAO 4. 3 is the current supported release. User demand remains strong: Dec 15 -Aug 31 : 1473 downloads (core+tools, mostly with Sherpa/Chips too) Linux Mac Solaris Source build 825 (of which about half were 64 -bit) 516 (of which 45 were 32 -bit) 9 123 This period we also saw 8 downloads of the older release CIAO 4. 2 CIAO 3. 4 is no longer available and we haven't had any requests for it. Solaris downloads from Kyoto, IPAC, MIT Solaris still used internally for some mission-critical Data System (e. g. database) uses but could save some effort by not packaging it for release * Should we retire the Solaris CIAO release? CIAO 4. 4 (Dec 2011) to be the last support for Solaris? CUC Oct 2011 CXC-SDS

Releases • CIAO 4. 3 released Dec 2010 • CIAO 4. 3. 1 patch to 'mkrmf' (problem with -100 C 1999 data) • CIAO 4. 4 release planned for Dec 2011 – maintenance/incremental improvements – new afterglow tool, new Sherpa functions, Ch. IPS GUI • CXCDS internal pipeline releases, including DS 8. 4 (Jun 28) – Aspect pipeline periscope correction – Subpixel correction now is default for ACIS TIMED – HRC dead time correction modified – Improved ACIS/LETG extraction – Celldetect not run any more – FITS keywords updated to match new standard REPRO 4 - new uniform reprocessing of archive now in prep. Subpixel, afterglow, BI serial CTI, Temp-dept. CTI, aspect, grating order sorting, header keyword consistency CUC Oct 2011 CXC-SDS

Community outreach • Supported booths at June AAS (Boston) • SDS scientists attended HEAD meeting • Organized X-ray Astronomy School and CIAO Workshop in August CUC Oct 2011 CXC-SDS

X-ray Astronomy School 2011

X-ray Astronomy School • • • • 2011 X-ray Astronomy School held at Cf. A in August 31 students from US, Italy, Spain, Canada (mostly grad students and postdocs) Sponsored by NASA, CXC and ESA-XMM http: //cxc. cfa. harvard. edu/xrayschool/ Lecturers from Cf. A, MIT, JHU, GSFC and MSFC (SDS: Aneta, Jonathan, Mike N, Dave H) Provided pre-prepared hands-on projects (although most students brought their own analysis problems) Support and tutoring during hands-on analysis sessions (SDS and DS) Very positive feedback Students focussed on Chandra data more than in previous XAS at GSFC XAS lectures focus on X-ray science and generic X-ray data analysis XAS was followed by 1 -day (Saturday) 8 th CIAO Workshop Similar format (talks followed by hands-on session) Focus on CIAO and Chandra data analysis Kudos to Liz Galle for a great job organizing CUC Oct 2011 CXC-SDS

Helpdesk • Helpdesk: 271 new tickets (Oct 25 – Aug 31), 3 still open • – – » » Median time to first answer 1. 7 hours (longest 7 days, waiting for user to provide sample files and for input from scientists) Median time to final answer 5 hours, longest this time was 60 days (followup questions from user) Answers generated 20 new docs, 21 new bugs, 15 RFEs what were the bugs? examples: - bug in Sherpa PSF fitting » - bug in mkrmf (EBOUNDS handled wrong for -110 C files in CIAO 4. 3, now patched) » - bug in aprates for nonstandard choice of parameters – – – CUC Oct 2011 63% did not require scientist or DS support 5 open tickets from previous period all now closed We also get direct contacts to SDS scientists outside the helpdesk system, at the rate of several per week CXC-SDS

CIAO Documentation • Using Google Analytics on the websites to track usage. Total number of unique visitors (as determined by IP address) from Mar 1 to Aug 31: CIAO - 13400 Ch. IPS - 1000 Sherpa - 3600 Chart - 600 • Tested, documented and updated reports on known software bugs, advertised bug updates • Documented DS releases, CALDB releases and CIAO script updates (release notes, threads, why topics); updated CSC pages • Improved various threads and explanations following user feedback • Documented the Cha. RT ray trace interface update • Reorganized and improved the imaging spectroscopy documentation, simplifying threads • Released a data caveat on the recent aimpoint shift • New 'Sherpa Snapshot' page for new Sherpa users; 2 new Sherpa threads, 3 updated, updated 'Python resources' page CUC Oct 2011 CXC-SDS

CIAO Docs – update to Web infrastructure T Ran link-checking script on all CIAO component web sites Corrected broken, dead links Added check for javascript errors on CIAO download page as part of site-wide security review Redesigned code for CIAO bugs pages Created an RSS feed for the CIAO news page Went to CXC-wide include files for header and footer code (improved maintenance) Revised software and calibration sections in thread synopsis and improved tagging of software-change-related information for easy searching

Sherpa improvements • • Sherpa is our multipurpose parameter-fitting application Improvements to Performance and Optimization – (ongoing by DS) Independent Python build - has been made available as a contributed package. Mac OSX version includes dmg image files for easy installation http: //cxc. harvard. edu/contrib/sherpa – 83 downloads of the standalone Sherpa (Jan 1 – Oct 5) Attended Randall Smith's mini-symposium on 'X-rays from Non-Equilibrium Plasmas' and participated in discussions between XSPEC, XSTAR, ATOMDB and Sherpa groups on the future of atomic models and their application to the data Beginning plans for a python library with flexible access functions to ATOMDB, providing more advanced capabilities to the user – can currently run ATOMDB via XSPEC support within Sherpa. Internal XSPEC support upgraded to XSPEC 12. 7 Advanced capabilities – see later ADS papers citing Sherpa - 352 CXC Quarterly Report Sep 2010 CXC-SDS

2 Making Data Analysis Easier • • 2. 1 2. 2 2. 3 2. 4 CUC Oct 2011 Scripts Statistics utilities Ch. IPS GUI TGCAT CXC-SDS

SDS Contributed Scripts We continue our effort to simplify Chandra analysis with improved threads and with scripts that wrap up the threads for simple use cases. Over the past few years we have been working our way through the analysis process from getting the s/w and data to analysing it. . . - ciao_install (CXCDS script) made getting CIAO much easier download_chandra_obsid gets you data if you know the obsid chandra_repro runs the thread to reprocess and get the updated evt 2 file fluximage (new) makes a set of exposure-corrected image in multiple energy bands – e. g. fluximage evt 2. fits out=outdir bin=4 bands=CSC - specextract makes PHA and response products - combine_spectra coadds multiple ACIS spectra and responses Next step is to make it easier to get a flux for a source by wrapping the aprates, modelfux, eff 2 evt functionality, and to make it easier to merge observations with WCS alignment. CUC Oct 2011 CXC-SDS

SDS Contributed Scripts Six releases of the CIAO contributed scripts package since March 2011. These are Python scripts written by SDS members for specific scientific tasks (see http: //cxc. harvard. edu/ciao/download/scripts/). • Positive feedback in helpdesk and email about some of the new scripts, especially specextract, chandra_repro and fluximage. Examples of new scripts: • • • fluximage: creates an exposure-corrected image for a single observation in one or more bands. Replaces use of merge_all for this case. This script was partly written as a component for developing merge_all 2, a forthcoming merge_all replacement check_ciao_version: report any packages that need to be updated – helper script which supports bug reporting. get_fov_limits: where is the data in my image? CUC Oct 2011 CXC-SDS

SDS Contributed Scripts - releases Full release content (see http: //cxc. harvard. edu/ciao/download/scripts/history. html for details). Contains new and updated scripts; releases typically monthly Examples • 4 Apr 2011, chandra_repro fixed to work around DM bug, handle CC mode right; specextract enhanced to allow user control of bkg. responses and to support RA/Dec regions (work around arfcorr limitation) • 25 Apr 2011, added get_fov_limits script – helps find good binning parameters for an image • 7 Jul 2011, new fluximage script, and improve specextract to handle -110 C data CUC Oct 2011 CXC-SDS

SDS Contributed Scripts: In Prep • • Preparing new merge_all 2 script. • The old merge_all is fine for cases where the observations are cocentered and close in time (same calibration) – e. g. recent split obsids The final version of the script will handle multiple energy bands, create fluxed images (as in the fluximage script) and will have the capability to automatically align observations using wcs_update We will release an initial version that replicates, with improvements, the scope of old merge_all (no WCS corrections, just reprojection) in the near future and phase in additional capabilities as resources allow. • • This script will fix a longstanding bug in which exposure maps are not correctly calculated. It does this by calculating exposure maps for each observation rather than for the merged events, and is therefore significantly slower CXC Quarterly Report Sep 2010 CXC-SDS

Statistics utilities • • • Goal – make it easier to interpret results of fitting New python statistics functions Callable in Sherpa CXC Quarterly Report Sep 2010 CXC-SDS

Sherpa – Likelihood Ratio Test (example of new stats functions) Is the addition of a gaussian line significant? Apply LRT to compare power law (p 1) with power law + line (p 1+g 1) Answer: p = 0. 016 Significant at 1. 6 % Level Black line: distribution of ratio of (likelihood with line)/(likelihood with no line) for ensemble of simulations Orange line: same for actual data CXC Quarterly Report Sep 2010 CXC-SDS

Chi. PS GUI – making plotting easier Ch. IPS GUI – under development by DS after SDS prototyping. SDS testing and feedback underway. Makes existing power of Ch. IPS much more accessible to users.

Full GUI panel includes terminal emulator (bottom), panel showing relationship of plot components (left). Use “View” menu to control which things appear.

TGCAT We have previously reported on the TGCAT grating data archive TGCAT continues to be kept up to date as new grating data go public – currently 1201 Obsids (833 HETG, 121 ACIS/LETG, 247 HRC/LETG) for 358 distinct sources Planned enhancement: ability to save plot configuration and apply it to a list of extractions (efficient way to compare many spectra) TGCAT queries: 6601 from 454 ip addresses (1 Mar -31 Aug) (VO 59% Web 31% CLI 10%) 42 package and 996 single-file downloads total 5. 8 Gb

3. Instrument/Observatory/Cal changes • • 3. 1 3. 2 3. 3 3. 4 CUC Oct 2011 PSF artifact study HRC blur calibration Handling ACIS afterglows Grating CC mode calibration CXC-SDS

Followup study of PSF Asymmetry Reported artifact (asymmetry) in PSF at previous quarterlies (at 0. 7”, few percent of flux) Now investigated data from 1999 -2010 Found PSF artifact was already there in Oct 1999 (Obsid 1385) although closer to PSF center (0. 5”+/-0. 1”) and at slightly different PA Dec 2010 AR Lac data suggests artifact stable since Dec 2000 HRMA modelling required to identify cause (post-launch shift, shell misalignment, ? ) Deconvolved images of AR Lac 0. 3” Core 0. 7” sector (artifact) 1” out, on left (“blob”) Left: 2010 Dec Right : 1999 Oct

HRC-I blur factor study Studied HRC-I instrumental blur: How do you need to blur an SAOTrace HRMA PSF to match HRC-I data? Used AR Lac data since 1999, including high S/N (18 ks) Dec 2010 obsn. Result: Blur factor 0. 18” to 0. 22” for r < 4 HRC-I pixels (r < 0. 5”) Blur factor 0. 25” for r = 4 – 10 pixels (r = 0. 5” to 1. 3”) Black line shows data Colored lines are PSF models with different gaussian blurs Recommend: blur with sigma= 0. 2” for most data. Extra blur beyond 0. 5” could be due to artifact? Will document for users

Afterglows A couple of years ago we found that our current afterglow treatment in the pipeline misses some afterglows that can show up as faint (~7 count) (false) sources The new acis_find_afterglow tool continues to undergo extensive testing. It is doing a good job identifying afterglows. We are currently testing the behaviour for identifying hot pixels – testing is looking good but is not yet complete. Improvements: - added extra parameters: cntthresh - minimum events in an afterglow nfpixreg, nfrepeat, tolerance - iterative calculation for nominal fluence probthresh - significance threshold for potential afterglow, 3 sigma = 0. 001 regwidth - size of reference region - improved understanding of computation of median fluence and its standard deviation - identified bugs in afterglow code and tested fixes Testing: explored set of 10 Obsids including all ten CCDs, with 22 M events and 10 k glows Explored full algorithm parameter space and confirmed that afterglow part of the tool now is performing to spec. Determined that default parameters are a robust choice (changing parameters does not significantly improve detection efficiency). Hot pixel detection: modified spec to clarify the definition of a hot pixel, the computation of the number of events on a possible hot pixel and of its significance. Analysed existing bad pixel files and discovered that ACIS S 4 streaks are sometimes identified as hot pixels; changed pipeline settings for destreak and modified pipeline to use it before afterglow detection.

CC mode analysis research Goals: - improve CC-mode HETG order sorting - develop method to determine HETG CC-mode background spectrum * Default CC mode order sorting window is wide to allow for gain variations * Because we collapse to 1 D, background from instrument is enhanced * Developed prototype code to support ACIS Cal team development of CC-mode CTI correction - input user-created order sorting window, event file, PHA-based background file Cyg X-1 current (black) narrow window (blue) b/g subtracted (orange) Lambda, order plot Narrow window (blue) and background spectrum (orange)

4. Improving algorithms and data quality • • 4. 1 4. 2 4. 3 4. 4 CUC Oct 2011 Improved PSF simulation: MARX enhancements Improved PSF simulation: Sub-pixel in MARX Improved grating processing: zero order location HRC dead-time factor – improved calculation (not discussed) CXC-SDS

PSF tools and analysis Cha. RT is the Chandra Ray Trace web page. It is a web interface to the CXC Optics Group's SAOTrace program. SDS supported release of the new Cha. RT in April (revised SAOTrace wrapper for internal use in the web service) We tested it against both old Cha. RT and against the HRMA team's standalone SAOTrace. MARX is the (MIT SDS developed) simplified, portable ray trace. We are enhancing it to be more flexible and more accurate. Cha. RT and MARX are both being used to investigate properties of the PSF by comparing with real data. .

MARX improvements: extended chip option Standard behaviour: apply chip gaps for realistic detector PSFs New optional behaviour: do not discard off-chip photons, suitable for calculation of PSFs for interpolation, encircled energies, etc.

MARX improvements: subpixel support MARX knows the exact ray-traced position of an event within a pixel. We can simulate the effect of lost information due to ACIS pixelization and reconstruct the PSFs that should be observed for different processing choices (pixel center, pixel randomization and EDSER sub-pixel reconstruction. Here we show the profiles and encircled energies for simulated PSFs with these different methods. Previous versions of MARX used the 'randomize' method.

Grating zero order Zero order can be distorted by pileup for bright sources Need zero order centroid for accurate grating coordinates (wavelength scale) Existing 'findzo' script isolates MEG and readout streak and fits their intersection – run if deemed necessary during V&V review In cases where even this fails, use a priori source position. Results of this in V&V to date: HETG/ACIS-S LETG-ACIS-s Total 833 121 tgdetect 65% 69% findzo 23% 27% a priori 12% 3%

Grating zero order Plan to include this algorithm within “tgdetect” pipeline/user tool - then can choose to run this instead of current celldetect method used by tgdetect. Second phase (2012) will be to review current data and develop heuristic for which method to use – will allow us to automate the choice in many (but not all) cases Plot shows that the need for findzo kicks in when the MEG count rate is about 0. 7 c/s.

Chandra Source Catalog: SDS Support 5. Chandra Source Catalog: SDS Support - Covered in Ian's talk Paper 2 (Characterization: Primini et al 2011) published as 2011 Ap. JS. . 194. . . 37 P - Supporting algorithm development for multi-obsid detect - Developed improved aperture photometry algorithm for crowded field case - Bayesian formalism to support contamination from nearby sources, allow informative priors for source intensity (eg. log N log S), improved (more robust and faster) implementation - Paper 3 (aperture photometry) in preparation - Limiting Sensitivity algorithm development – now using a likelihood approach based on C-statistic in the MLE source fit, approch similar to 2 XMM's - Simulation code for multiple obsids, instrumented to tag events for each source; event tagging within MARX as a new standard feature. - Catalog threads: new thread on Proposer website for 'finding a point source in CSC' http: //cxc. cfa. harvard. edu/proposer/threads/csc CXC Quarterly Report Sep 2010 CXC-SDS

6. Sherpa advanced capabilities • • • Flexible fitting with templates (1 D and 2 D) – iterating with DS on specs and development, investigating fitting techniques, collecting example applications to real data. Supports fitting multiwavelength SEDs combined with X-ray data Statistics functions including goodness-of-fit simulations with Cash, ppp hypothesis test and improvements to flux error simulations – DS provided to SDS for evaluation and test in mid-Sep. Testing performance and behaviour of pyblocxs statistics code using simulations (work continuing) CXC Quarterly Report Sep 2010 CXC-SDS